Carbohydrate regulation of hepatic gene expression. Evidence against a role for the upstream stimulatory factor.

Hepatic expression of the genes encoding L-type pyruvate kinase (L-PK) and S14 is induced in rats upon feeding them a high carbohydrate, low fat diet. A carbohydrate response element (ChoRE) containing two CACGTG-type E boxes has been mapped in the 5′-flanking region of both of these genes. The nature of the ChoRE suggests that a member of the basic/helix-loop-helix/leucine zipper family of proteins may be responsible for mediating the response to carbohydrate. Indeed, the upstream stimulatory factor (USF), a ubiquitous basic/helix-loop-helix/leucine zipper protein, is present in hepatic nuclear extracts and binds to the ChoREs of L-PK and S14 in vitro We have conducted experiments to determine whether USF is involved in the carbohydrate-mediated regulation of L-PK and S14. For this purpose, dominant negative forms of USF that are capable of heterodimerizing with endogenous USF but not of binding to DNA were expressed in primary hepatocytes. Expression of these forms did not block either S14 or L-PK induction by glucose. In addition, we have constructed mutant ChoREs that retain their carbohydrate responsiveness but have lost the ability to bind USF. Together, these data suggest that USF is not the carbohydrate-responsive factor that stimulates S14 and L-PK expression and that a distinct hepatic factor is likely to be responsible for the transcriptional response.

The mammalian liver is capable of converting excess dietary carbohydrate into triglycerides for storage. Feeding of a diet high in simple carbohydrates and low in fats results in the induction of many hepatic enzymes involved in lipogenesis, including liver-type pyruvate kinase (L-PK), 1 fatty acid synthase, malic enzyme, and acetyl-CoA carboxylase (for reviews see Refs. 1 and 2). In each case, the increased enzyme production correlates with an increase in mRNA levels, and in some of these cases the increase in mRNA has been shown to be due to elevated transcription of the corresponding genes (3)(4)(5)(6)(7)(8). Little is known about the molecular mechanisms responsible for this transcriptional response.
In our efforts to characterize the regulation of lipogenic gene expression, we have been studying the genes encoding L-PK and S 14 in rat liver. These genes are transcriptionally induced upon carbohydrate feeding (3,9,10). This induction can be reproduced in cultured primary hepatocytes by varying the glucose concentrations in the media (9,11,12). Increased glucose metabolism is thought to generate the intracellular signal in the regulatory pathway. Although the exact nature of this pathway remains unknown, several possible intracellular signals have been proposed (13)(14)(15). While insulin is necessary for the regulation, it has been shown to play only a permissive role in promoting effective glucose metabolism (16,17). Sequences responsible for mediating the carbohydrate response have been mapped in the 5Ј-flanking region of both the S 14 and L-PK genes (18 -22). Both genes contain a common regulatory element that consists of two 5Ј-CACGTG-type E box motifs separated by 5 base pairs (19,23). Linking multiple copies of this element to a basal promoter is sufficient to induce carbohydrate-responsive activity; this element is thus referred to as a carbohydrate response element (ChoRE). However, in the context of the natural promoter, each gene contains a distinct accessory factor site adjacent to the ChoRE that is necessary for the full extent of the response (22)(23)(24).
The presence of the CACGTG-like sites within the ChoRE suggests that a member of the basic/helix-loop-helix/leucine zipper (b/HLH/LZ) family of transcription factors may be involved in the carbohydrate-mediated regulation (25,26). These proteins all contain a basic DNA binding region followed by helix-loop-helix and leucine zipper dimerization domains. The b/HLH/LZ class includes Myc and its related family members, along with other proteins such as TFE3, TFEB, SREBP1/ ADD1, and the upstream stimulatory factor (USF) (27)(28)(29)(30)(31)(32)(33). Of these, USF appears to be the predominant b/HLH/LZ factor in hepatic nuclear extracts. USF was first identified by its ability to bind to the upstream stimulatory element of the adenovirus major late promoter (34,35). In vitro electrophoretic mobility shift assays (EMSAs) using hepatic nuclear extracts produced one major shifted complex that bound specifically to the S 14 and L-PK ChoREs and which was "supershifted" by the addition of an antibody against USF (21,22,24). However, the USF binding site from the adenovirus major late promoter was unable to substitute for the ChoRE in either the L-PK or the S 14 gene, indicating that USF alone is not capable of mediating the response (21). Additionally, the ubiquitous nature of USF (36), coupled with its ability to interact with sequences in other genes that are not regulated by glucose (37)(38)(39), raised the question as to how the specificity of the response is achieved. We thus set out to determine whether or not USF is involved in the regulatory pathway induced by carbohydrate.

Primary Hepatocyte Culture and Transfection
The procedure previously described was followed (20). Primary hepatocytes were isolated from male Sprague Dawley rats (180 -240 g) maintained on a 12-h light/dark cycle with free access to normal chow. Following a 3-6-h attachment period, cells were transfected using the Lipofectin reagent (Life Technologies, Inc.) in modified Williams E medium (lacking methyl linoleate and glucose) supplemented with 23 mM HEPES, 0.01 M dexamethasone, 0.1 unit/ml insulin, 1 unit/ml penicillin, 1 g/ml streptomycin, and 11 mM glucose for 12-14 h. Cells were subsequently incubated for 48 h in medium containing either 5.5 or 27.5 mM glucose and harvested for chloramphenicol acetyltransferase (CAT) assay. For studies using dominant negative forms of USF, Matrigel (Collaborative Biomedical Products) was added to the cells at a concentration of 0.5 mg/ml after the overnight transfection. This treatment has been shown to increase the time that cultured hepatocytes remain responsive to elevated glucose concentrations (40). After the addition of Matrigel, plates were incubated in low glucose medium for 24 h before the 48-h incubation with low or high glucose. Percentage conversion of chloramphenicol to its acetylated forms was determined by thin layer chromatography followed by phosphor screen autoradiography, with subsequent scanning and quantitation (Molecular Dynamics, Inc.)

Plasmid Constructs and Oligonucleotides
S 14 Mutants-The pS 14 CAT(An)-4316-mut1, -mut2, and -mut3 constructs were prepared using PCR mutagenesis to engineer mutations at the ChoRE E boxes and the accessory factor site (41). The exact location of each mutation is shown in Fig. 1A (Mut1, Mut2, and Mut3). Two primers were used for each construct, one directed upstream from the mutation site and one directed downstream. These primer pairs contained either an XbaI site or a BamHI site in place of the motif to be mutated. The polymerase chain reaction was carried out using pS 14 CAT(An)-4316 (9) as a template. PCR products were digested with HindIII and XbaI/BamHI or XbaI/BamHI and SfiI to generate fragments extending from Ϫ2111 through the mutation and from the mutation to Ϫ767, respectively. These fragments were then ligated into the HindIII and SfiI sites of pS 14 CAT(An)-4316. Sequences of the mutations were confirmed by DNA sequencing.
USF Expression Vectors-Plasmids psvUSF2 and psvUSF2⌬B, which encode mouse USF2 and USF2⌬b (39), were generously provided by Dr. Michéle Sawadogo (M. D. Anderson Cancer Center, University of Texas). USF2 coding regions were transferred into the CMV4 plasmid (42). This was accomplished by ligating a BglII linker to the EcoRI site and inserting the desired USF2 fragment into the BglII site of CMV4. A similar construct containing USF2⌬N⌬b was prepared by carrying out the polymerase chain reaction on the USF2 construct using primers with appended MluI or XbaI sites to generate a product encoding USF2 amino acids 1-6 and 249 -346. The PCR product was cleaved with MluI and XbaI and ligated into CMV4. A CMV/USF1 expression plasmid was provided by Drs. Philippe Pognonec and Robert Roeder (Rockefeller University). The bAP4/USF1 construct was prepared by using PCR to amplify the N-terminal and C-terminal portions of USF1 separately. Internal primers were directed upstream and downstream from either side of the basic region; each had an extended overhang encoding the AP4 basic region containing an NspV site. The external primers contained appended HindIII or XbaI sites. PCR products were digested with HindIII and NspV or NspV and XbaI and inserted into the HindIII and XbaI sites of CMV4 by triple ligation. The USF2/VP16 construct was made by first amplifying the entire USF2 sequence to eliminate the stop codon and replace it with a BglII site. The PCR product was cleaved with HindIII and BglII and inserted into pSJT1193, a plasmid containing the C-terminal activation domain of VP16, which was kindly donated by Dr. Steve Triezenberg (Michigan State University). The USF2/VP16 fusion was then removed from this plasmid by digestion with KpnI and HindIII and inserted into CMV4.
ChoRE Mutations-The parental PK(Ϫ96)CAT plasmid into which altered ChoRE sequences were inserted has been described (18). Oligonucleotides were synthesized containing the sequences shown in Fig. 4 flanked on each side by a HindIII site. These fragments were cloned into the unique HindIII site in PK(Ϫ96)CAT just upstream of the L-PK basal promoter. Plasmids were subjected to sequence analysis to determine the orientation and copy number of the insert.

Nuclear Extract Preparation and EMSAs
Rats were fasted overnight and refed a high carbohydrate, low fat diet (ICN Biomedicals, Inc.) for 4 h before sacrifice. Nuclear extracts were prepared by the procedure of Lavery et al. (43). EMSAs were performed essentially as described (20). A typical reaction contained 50,000 cpm of 32 P-labeled oligonucleotide (0.5-1 ng) and 3 g of nuclear extract. After 45 min at 23°C, samples were subjected to electrophoresis on 4.5% nondenaturing polyacrylamide gels in a Tris-glycine buffering system (20). For competition experiments, unlabeled oligonucleotides were added to the radiolabeled probe prior to the addition of nuclear extract.

Roles of the ChoRE and Accessory Site in Carbohydrate
Regulation of the S 14 Gene-The sequences of the rat S 14 gene involved in supporting a transcriptional response to carbohydrate were previously mapped to the 5Ј-flanking region between Ϫ1467 and Ϫ1422 (23). This region corresponds to a liver-specific DNase I hypersensitive site that appears developmentally at the time the S 14 gene is first expressed (44). Within this regulatory region, the ChoRE was localized to the sequences between Ϫ1448 and Ϫ1422, consisting of two CACGTG-type E box motifs; the 5Ј motif is a perfect CACGTG, while the 3Ј CCTGTG site is a 4/6 match (Fig. 1A). In addition, an adjacent accessory factor site was localized between Ϫ1467 and Ϫ1448. This site did not support a response to carbohydrate on its own but greatly augmented the response of the S 14 ChoRE to glucose. The nature of the factor hypothesized to bind to this accessory site is under investigation. Mapping of the S 14 regulatory sites utilized plasmids in which the elements were linked directly to a basal promoter, with large segments of the S 14 5Ј-flanking region deleted. We therefore felt it was important to determine the role of these sites in the context of the natural S 14 gene. Plasmids containing S 14 sequences from Ϫ4316 to ϩ18 were linked to the CAT reporter gene. These plasmids contained clustered mutations within the 5Ј E box, the 3Ј E box, or the accessory factor site. The constructs were transiently transfected into primary rat hepatocytes, which were cultured in low (5.5 mM) or high (27.5 mM) glucose conditions for 48 h before being harvested and assayed for CAT activity. As depicted in Fig. 1B, mutations at each of these sites dramatically decreased the glucose response, as compared with the response of the wild type S 14 control plasmid. This experiment clearly demonstrated the necessity for all three of these motifs for supporting the full glucose response of the S 14 gene.
Dominant Negative Forms of USF Do Not Inhibit the Carbohydrate Response-Localization of the S 14 and L-PK ChoREs allowed us to investigate the identity of the carbohydrateresponsive factor. Since USF is capable of interacting with either of these ChoREs in vitro (21,22,45), we set out to determine whether it is necessary for the response in hepatocytes. To test this, dominant negative USF proteins were expressed in hepatocytes, and the effect on the glucose-mediated induction of the S 14 and L-PK promoters was examined. USF is present endogenously in two isoforms, designated USF1 and USF2, which are capable of binding to DNA as homodimers or as heterodimers with each other (33,46,47). Fig. 2 shows the various forms of USF used in our studies. The dominant negative protein bAP4/USF1 contains the basic region of the b/HLH/LZ protein AP4 in place of the USF1 basic region, thus altering its DNA binding specificity (48). USF2⌬b lacks the basic region altogether, while USF2⌬N⌬b lacks both the basic region and the N-terminal transactivation domain. Both of these proteins are therefore incapable of binding to DNA, and USF2⌬N⌬b further lacks transactivation ability. Since all three of these forms retain the helix-loop-helix and LZ regions, they should be capable of heterodimerizing with USF1 or USF2 but incapable of binding to the CACGTG motif.
Before testing these dominant negative forms of USF on the carbohydrate response, it was important to demonstrate their effectiveness. Although USF has been implicated in the expression of several hepatic genes (e.g. Refs. 37, 39, and 49), to date no genes have been shown to be directly induced by USF within the hepatocyte. In addition, cotransfection of USF into hepatocytes with reporter constructs containing USF binding sites has not led to transactivation. 2 To circumvent this difficulty, we constructed a "superactivator" chimeric USF protein in which the transcriptional activation domain of VP16 is linked to the C terminus of USF2 (Fig. 2). Introduction of a plasmid expressing this protein stimulated the activity of a reporter plasmid containing the adenovirus USF binding site in a dosedependent manner. 2 In addition, the USF2/VP16 construct stimulated expression of a plasmid containing two copies of the S 14 ChoRE linked to the L-PK basal promoter and CAT in low glucose conditions (Fig. 3A). Thus, USF has the ability to bind to the S 14   of the three dominant negative forms was capable of inhibiting the action of USF2/VP16 (Fig. 3B, stippled bars). This indicates that the dominant negative forms of USF are expressed and functional within the primary hepatocyte.
The effect of the three dominant negative forms of USF on carbohydrate induction was subsequently evaluated. For this purpose, we used a construct containing two copies of the S 14 ChoRE linked to the L-PK basal promoter in the absence of the accessory factor site (Fig. 3A), since this was the simplest glucose-responsive construct that we had found. Following transfection of the reporter gene construct with or without constructs expressing the dominant negative USF forms, cells were incubated for 24 h in low glucose conditions to allow for accumulation of the dominant negative proteins. Cells were then incubated in low or high glucose conditions for 48 h, at which time they were harvested and CAT assays were performed. None of the three dominant negative forms were able to inhibit the glucose response of the reporter construct (Fig.  3B). Similar experiments were carried out using a reporter plasmid containing the L-PK ChoRE. In this case, the L-PK ChoRE was tested in its natural context (Ϫ197 to ϩ12) with the accessory factor site (Fig. 3A). Again, we found no significant effect of expressing the dominant negative USF forms on carbohydrate induction (Fig. 3C). Finally, we observed no effect of the dominant negative USF proteins on the glucose response of a plasmid containing sequences from Ϫ4316 to ϩ18 of the S 14 gene (data not shown). These experiments suggested that USF is not involved in the carbohydrate induction of the S 14 or L-PK promoter.
Dissociation of Sequence Requirements for ChoRE Activity and USF Binding-The results of the experiments with dominant negative forms of USF were in contradiction to a published report by Lefrançois-Martinez et al. (50) studying L-PK gene induction in a hepatoma cell line. We therefore assessed the potential role of USF by a second criterion. For this purpose, we examined the bases of the ChoRE sequence that are necessary to confer a response to glucose. The goal of this study was to determine whether mutations that affect function also affect USF binding. As a starting point, we used an optimized ChoRE containing two perfect CACGTG motifs separated by 5 base pairs in the context of the S 14 ChoRE sequences. Since this element is capable of supporting a glucose response in the absence of an accessory site, we thought that interpreting the effects of mutations should be more straightforward. We then altered individual bases in each CACGTG motif to create palindromic mutations. The altered ChoRE sequences are shown in Fig. 4. These elements were cloned into a CAT reporter vector upstream of the L-PK basal promoter. Following transfection and a 48-h incubation in low or high glucose medium, cells were harvested for CAT assays. Mutation of positions 1-4 in each E box motif completely abolished the carbohydratemediated regulation (Fig. 5). Motifs containing mutations at the fifth or sixth position, however, were still able to confer a response to glucose. We also prepared plasmids in which the altered ChoREs were placed adjacent to the L-PK accessory factor site to test their effects in a context more similar to the natural situation. The response pattern of this series of plasmids, however, paralleled that of the constructs with only the basal promoter. 3 Thus, at least with the specific base changes tested, the first four bases of the E boxes were more critical for the carbohydrate response than the last two bases.
One of the more striking results of these experiments came from the plasmid containing mutations at position 5 of each motif. The mutations created two CACGgG sites (lowercase letters here and elsewhere represent differences from the consensus CACGTG motif), identical to those found in the natural L-PK ChoRE, but flanked by bases surrounding the S 14 ChoRE. This plasmid consistently responded strongly to glucose treatment, with a -fold induction approximately twice as high as that of the plasmid containing two perfect CACGTG motifs. Perplexingly, mutational studies of the L-PK promoter had demonstrated that the ChoRE, in the absence of the accessory factor binding site, was not capable of supporting a response (19,22). Two possibilities might explain these results. The first is that the minor difference in the positioning of the ChoRE relative to the L-PK basal promoter might dramatically affect the glucose response. The second is that the nature of the bases surrounding and separating the E box motifs could modulate activity. To distinguish between these possibilities, we constructed a plasmid containing the natural L-PK ChoRE (mutant 5(PK); Fig. 4) in the same context as the other plasmids used in these experiments. This plasmid supported only a 2-fold glucose response (Fig. 5), suggesting that the bases flanking and separating the E boxes are indeed relevant to the response level.
An additional sequence was also tested in these studies. capable of binding to b/HLH/LZ proteins such as Myc and Max, but was unable to interact with USF. We therefore felt that this sequence might be useful for assessing the role of USF and set out to determine whether it could function as a ChoRE. Transfection analysis of a plasmid containing two CAtGcG sites, again separated and flanked by S 14 bases, showed that this sequence could indeed support a glucose response approximately equal to that of the plasmid containing two CACGTG motifs (Fig. 5).
Following these functional studies, we used two types of experiments to test the ability of the mutant ChoREs to bind to USF. The first involved cotransfecting primary hepatocytes with reporter plasmids containing the various mutant ChoREs and the expression plasmid encoding the USF2/VP16 chimeric protein. In this way, we could evaluate the binding of USF under conditions present in the hepatocyte. As expected, the USF2/VP16 protein was capable of stimulating expression from the construct containing two "perfect" CACGTG motifs (Fig. 6). Constructs containing mutations at E box positions 1-4, which did not respond to glucose, were also not responsive to USF/ VP16 ( Fig. 6 and data not shown). Interestingly, expression of those plasmids containing altered ChoREs that were glucoseresponsive was also not stimulated by USF/VP16. This separation of the carbohydrate response from activation by USF suggests that USF is not involved in the glucose-mediated regulation.
A second approach to test the role of USF in carbohydrate induction was to use in vitro EMSAs to determine the ability of the altered ChoREs to bind USF. Mobility shift assays were performed in which rat hepatic nuclear extracts were incubated with the radiolabeled oligonucleotide containing two perfect CACGTG motifs. This probe gave one primary complex that comigrated with a complex formed by an oligonucleotide containing the adenovirus USF site (data not shown). This complex was supershifted completely by the addition of an antibody to USF. Furthermore, the band was identical in migration to that seen when bacterially expressed USF was used instead of nuclear extract. Thus, this complex represents binding of USF to the probe sequence. Some of the altered ChoREs were then selected for use as competitors to determine their relative binding affinities. For this study we used only those sequences that retained carbohydrate responsiveness, with the exception of one negative control (mutant 4) that did not respond. Fig. 7A depicts an example of a representative compe-tition experiment, while Fig. 7B demonstrates the quantified results of the overall study. The "self-competitor" containing two CACGTG sites competed quite strongly for the bound complex. Three of the other sequences showed lesser degrees of competition. The oligonucleotide containing the CACGgG motifs flanked by S 14 bases, which was highly responsive to glucose, competed approximately 20 -25 times less effectively than the CACGTG sequence. The competitor with the natural L-PK ChoRE and the competitor with the CACGTt motifs, both of which were somewhat less glucose-responsive, competed approximately 40 -50 times and 75-100 times less efficiently, respectively. The CAtGcG-and CACtTG-containing competitors were essentially unable to compete for binding. These results were reproduced when recombinant human USF1 was used in place of nuclear extract (data not shown). These experiments demonstrated a lack of correlation between carbohydrate-responsive activity and USF binding. DISCUSSION We have previously identified by deletional mutagenesis a carbohydrate response element between positions Ϫ1448 and Ϫ1422 in the 5Ј-flanking region of the S 14 gene (23). This sequence is similar to the ChoRE of the L-PK gene (also called the glucose/insulin response element (19)). In either case, two CACGTG-like E boxes separated by 5 base pairs are present. In the case of the S 14 ChoRE, however, the 3Ј CCTGTG motif has only a 4/6 match with the consensus sequence. Two lines of evidence support the importance of this motif. First, mutations introduced into the 3Ј E box that make it a better fit to the consensus CACGTG resulted in stronger responses to glucose (23). Second, we have now shown that specific mutagenesis of the 3Ј E box in the context of the natural S 14 gene largely abolished the ability of the S 14 promoter to respond to glucose. Likewise, mutation of the 5Ј E box eliminated the response. Previous work has shown that both E boxes of the L-PK gene are also critical for control (19,22). Thus, the presence of two E box motifs is a critical feature for mediating the glucose response of these two members of the lipogenic enzyme family.
The S 14 and L-PK ChoREs are sufficient to support a response to glucose when linked in multiple copies to a basal promoter or mutated to give an optimal binding site. However, in either case an adjacent accessory factor site is required for the full extent of the response in the context of the natural gene. For the L-PK promoter, the hepatic-enriched factor HNF-4 is capable of binding just 3Ј to the ChoRE and greatly enhancing its activity (22,24). In the case of the S 14 gene, the accessory site located immediately 5Ј to the ChoRE binds an as yet unidentified nuclear factor (23). Mutating this element in the context of the natural S 14 gene greatly diminished the glucose response. Hence, this site appears to play an important role in supporting regulation. A similar situation has been found in several other genes subject to transcriptional regulation by hormonal and metabolic factors. For example, the gene encoding the low density lipoprotein receptor is activated when sterol levels are low by the action of the transcriptional factor SREBP (52). The action of SREBP by itself is weak, but it is greatly enhanced by binding of the ubiquitous factor Sp1 to adjacent sites within this gene (53,54). Sp1 also synergizes with SREBP in the activation of the fatty acid synthase and acetyl-coenzyme A carboxylase promoters (55,56). Interestingly, Daniel and Kim (57) have recently reported that Sp1 binding is also required for glucose activation of the acetylcoenzyme A carboxylase promoter. Whether this Sp1 binding site is serving as an accessory factor to an undetected ChoRE or whether it is the binding site for the actual carbohydrateresponsive factor remains to be determined. Neither the S 14 nor the L-PK promoter appears to contain binding sites for Sp1.
The presence of CACGTG-type E boxes within the ChoRE strongly suggests that a member of the b/HLH/LZ family of transcription factors is involved in control. In particular, USF has emerged from this family as a candidate for the carbohydrate-responsive factor, based largely on its ability to bind to the S 14 or L-PK ChoREs in vitro (22,45). However, we have found no evidence that USF is involved in the glucose regulation of the L-PK or S 14 genes in vivo. The expression of three different dominant negative forms of USF did not affect the carbohydrate response of either ChoRE in primary hepatocytes, although the same proteins were able to block the action of cotransfected USF2/VP16. This finding is in direct contradiction with that of Lefrançois-Martinez et al. (50). This group found that a dominant negative form of USF1 similar to USF2⌬N⌬b could block the L-PK glucose response in the glucose-responsive, hepatocyte-derived cell line mhAT3F. It is difficult at present to reconcile this contradiction. Several differences in the experimental strategies might be important. The use of the mhAT3F cell line versus primary hepatocytes constitutes a significant dissimilarity. While the extent of the differences between these two cells is unclear, they are certainly not identical. For example, as opposed to the primary hepatocyte, the glucose response in the mhAT3F cell is insulin-independent. This is apparently due to the fact that a hexokinase other than glucokinase is responsible for initiating the glycolytic process. Although not specifically known, it is also possible that the mhAT3F cells could contain less endogenous USF than the primary hepatocyte. If this were true, less dominant negative USF would be needed to inhibit the endogenous USF. Another significant difference in the experimental protocols is the time between transfection of the dominant negative USF forms and initiation of the glucose induction period. While we used 24 h in the present study, Lefrançois-Martinez et al. allowed a 72-h period. It may be that 72 h is needed to effectively sequester the endogenous USF pool and thus inhibit USF action. On the other hand, it is also conceivable that over 72 h the effective suppression of USF could lead to reduced expression of another transcription factor (e.g. HNF-1 or HNF-4) that is required for transcription of the L-PK promoter. In this way, the action of the dominant negative form of USF on the glucose response would be secondary rather than direct.
Given these conflicting results, we sought an additional way in which the role of USF in the glucose response could be addressed. The correlation between USF binding and glucose responsiveness was examined in a set of ChoRE mutations. The identification of mutations that discriminate between USF binding and function argues against a role for USF. In particular, mutation of the fifth position of the CACGTG motifs to give CACGgG resulted in an increased response to glucose when tested in the context of S 14 ChoRE sequences. This mutation was not activated by USF2/VP16 in cotransfection experiments and bound USF less tightly than the element containing CACGTG motifs by EMSA. This decreased binding was observed both for nuclear extract, which predominantly consists of USF1/USF2 heterodimers (36,47), and recombinant USF1, which should be present as homodimers. Even more strikingly, alteration of each E box motif to CAtGcG resulted in a ChoRE that responded nearly as well as the CACGTG-containing ChoRE. As reported by Blackwell et al. (51), this sequence did not effectively bind USF either in cotransfection assays or by EMSA. Thus, a role of USF homodimers or heterodimers in glucose activation appears unlikely.
It is worth noting that previous work from both our laboratory and others has shown that USF can bind to the PK ChoRE sequence (22,45). This is consistent with the demonstration that USF activated expression of L-PK promoter constructs when cotransfected into mhAT3F cells. Our competition studies show, however, that this binding is of significantly lower affinity than that to the CACGTG motif. It is hard to reconcile a reduced binding affinity with a higher potential to activate transcription if USF is indeed the carbohydrate-responsive factor. While we cannot exclude the possibility that USF may function as a complex with another unidentified factor that alters its affinity for ChoRE sequences compared with USF alone, we consider this possibility to be less likely given that no other members of the b/HLH/LZ family have been found to heterodimerize with USF.
It is also worth noting that mutants 2 and 5 possess the exact same E box motifs but in different orientation relative to each other. These two constructs differ dramatically in their abilities to support a glucose response, implying that the relative orientation of the two E boxes is critical. To further explore this question, we have tested a construct in which the same E box motif is present in a direct repeat fashion: CACGgGnnnnnCACGgG. As with mutant 2, this construct was essentially incapable of supporting a glucose response. 3 Previously, we observed that the spacing between the ChoRE E boxes is also critical for regulation (23). Together, these results indicate that a highly specific interaction between two binding factors must occur for an effective glucose response.
If USF is not the carbohydrate-responsive factor, then why do we not see another unique factor capable of binding to the ChoRE sequences in vitro? At present, we speculate that such a factor or complex may be present in liver but at a much lower concentration than USF, such that detecting this complex in crude nuclear extracts of liver may not be possible. A similar situation was encountered in the search for SREBP, which binds to the sterol response element of the low density lipoprotein receptor gene. Although proteins are present in crude nuclear extracts that bind the sterol response element, none proved to be the functional factor mediating the response to sterol (58). Only upon significant purification was the physiologically significant SREBP detected. We are currently using the CAtGcG element ChoRE, which is not bound significantly by USF, as a tool to search for such a complex involved in carbohydrate regulation.